The search for exotic knot traces
Marc Kegel, Jonathan Spreer · 2026
Read The search for exotic knot traces by Marc Kegel, Jonathan Spreer (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-22438.txt.
Introduction
Kegel and Spreer study knot friends, pairs of distinct knots whose complements give the same filled manifold. They build a practical search algorithm for finding a friend of a given knot. The search runs on the SnapPy and Regina toolkits for three-manifold topology. With it, the authors compile a census of simple knots that admit friends. They then use those data to frame conjectures about when knot friends occur.
Catalog record
| CREATOR | Marc Kegel, Jonathan Spreer |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 67105 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.22438
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
The search for exotic knot traces Marc Kegel, Jonathan Spreer Abstract . Two distinct knots are said to be friends if their complements, filled along the 0 0 -slope, produce diffeomorphic 3 3 -manifolds. In this article, we develop a practical algorithm, implemented using SnapPy and Regina , to search for a friend of a given knot. As an application, we construct a census of simple knots that admit friends and use these data to formulate conjectures about knot friends. 2020 Mathematics Subject Classification 57R65; 57K10, 57K32 1. Introduction We say that two knots in the 3 -sphere S^{3} are friends if they have diffeomorphic 0 -surgeries. In the literature, there exist many examples of pairs of knots that are friends, see for example [ 43 , 42 , 48 , 40 ] . On the other hand, many simple knots (like the unknot, the trefoil, or the figure eight) are known to have no friends [ 25 , 44 , 10 ] . The study of friends is motivated by the observation that a pair of friends with different smooth sliceness status 1 1 1 Recall that a knot is smoothly slice if it bounds a smoothly embedded disk in the 4 -ball. yields a counterexample to the smooth 4 -dimensional Poincaré conjecture [ 34 , Problem 1.19] , cf. [ 22 , 40 ] . Apart from the possibility of using friends to disprove the smooth 4 -dimensional Poincaré conjecture, this article is motivated by the following questions. Question 1.1 Given a knot K , does K have a friend? 1.1 should be understood to describe general topological, algebraic, or geometric conditions on K that ensure the existence or non-existence of a friend. Question 1.2 Is there an algorithm that, given a knot K , returns true if K has a friend, and false otherwise? To make 1.2 more precise we can assume that the input knot K is given as a knot diagram, or as a triangulation of the exterior of K . In the former case, we often abuse notation and refer to K as both the knot and its representation as an input diagram. 1.1. Algorithmic search for friends The main contribution of this article is code and the results of running this code. Namely, we have the following result. Algorithm 1 (Friend search) There exists a practical and implemented algorithm that takes as input a diagram of a knot K , or a triangulation of the exterior of K , and outputs a (possibly empty) list of friends of K . Note that the solution of the homeomorphism problem for closed 3 -manifolds or 3 -manifolds with toroidal boundary [ 35 ] together with the Gordon–Luecke theorem [ 26 ] implies the existence of an algorithm that takes as input a knot K and outputs in finite time any given friend of K – if it exists: Enumerate all knots K^{\prime} by increasing crossing number, create their 0 -surgeries K^{\prime}(0) , and check if K(0) and K^{\prime}(0) are diffeomorphic. This approach was taken in [ 1 ] for knots of low complexity. However, since the enumeration of knots has super-exponential running time in the crossing number, this is impractical in general. Nonetheless, on a theoretical level, 1.2 reduces to certifying that a given knot has no friend. Algorithm 1 , on the other hand, takes a knot diagram of K and first constructs a 1 -vertex triangulation of its 0 -surgery. It then removes an embedded closed curve c from K(0) , and checks if K(0)\setminus c is the complement of a knot in S^{3} using the 6 -theorem [ 2 , 36 ] . This procedure is then iterated over all curves c in K(0) up to ambient isotopy (i.e. over all knots in K(0) ) using a sensibly chosen order. For producing a 1 -vertex triangulation of K(0) from the input diagram, we use readily available functionality of SnapPy [ 15 ] . For enumerating a list of curves c to remove from K(0) we use the fact that every edge in a 1 -vertex triangulation describes an embedded closed curve in the underlying manifold: We first produce a list of 1 -vertex triangulations of K(0) by exhaustively modifying the initial triangulation using local moves (e.g. bi-stellar flips [ 45 , 50 ] ). Then every edge of every modified triangulation is added to the collection of curves. Both, exhaustive enumeration of “neighbor triangulations”, as well as drilling out edges from a triangulation is readily available functionality of Regina [ 13 ] . We describe this algorithm in Section 2 . By standard arguments (carried out rigorously in a more specialised setting in [ 31 ] ), we can expect every ambient isotopy class of a closed curve in K(0) to be realised as an edge in some triangulation of K(0) . It follows that this approach will, eventually, find every given friend K^{\prime} of K . Worst case running times of this procedure quickly become excessive – even if K^{\prime} is observed by an edge in a triangulation of K(0) that is only a small number of local moves away from the initial triangulation of K(0) . In practice, however, the approach is efficient in many cases, as we discuss in Section 2 and 3 below. A similar algorithm for hyperbolic 0 -surgeries is described in [ 19 ] and further extended and used in [ 18 ] . There, the curves c are enumerated by considering paths in the 1 -skeleton of a dual cell-decomposition, or by drilling out short geodesics. 1.2. A census of friends We apply the friend search – Algorithm 1 – to standard datasets of knots: The low crossing number knots (i.e. prime knots that admit diagrams with at most 12 crossings) and the SnapPy census knots (i.e. the hyperbolic knots whose complements can be ideally triangulated by at most 9 ideal tetrahedra). Low-crossing number knots are described using the Dowker–Thistlethwaite (DT) notation [ 17 ] , as in the Hoste–Thistlethwaite–Weeks table [ 30 ] , cf. [ 38 ] . Census knots are described using the notation of Dunfield [ 20 ] as implemented in SnapPy [ 15 ] . For example, the figure eight knot is K4a1 in the DT notation and K2\_1 in the census knot notation. For more background on the SnapPy census knots, we refer to [ 21 , 4 , 9 , 7 , 6 , 37 ] and references therein. Theorem 3.1 (1) Among the 2977 low crossing number knots, at least 1867 have a friend. (2) Among the 1267 SnapPy census knots, at least 183 have a friend. The low crossing number knots that admit a friend are displayed in Table 4 . PD codes of potentially minimal diagrams of the friends can be accessed at [ 33 ] . Their average crossing number is 159 , while the median is 109 . The friend with a diagram of maximal crossing number is a friend of K12a951 and has 1259 crossings. For the SnapPy census knots, the average crossing number is 53 , the median is 42 , and the maximal crossing number diagram of a friend has